The selection of optimal afforestation models is critical for restoring soil functions in degraded mine lands, yet the long-term microbial mechanisms driving these processes remain poorly understood. This study aimed to assess the effects of five different forest restoration models on soil carbon (C) sequestration, mineralization, and the underlying microbial functional gene potential after 32 years of development at the Pingshuo opencast coal mine, a major mine on the eastern Loess Plateau, China. We compared pure stands of Ulmus pumila L. (UP), Pinus tabuliformis C. (PT), and Robinia pseudoacacia L. (RP), alongside a mixed broadleaf (Robinia-Ulmus-Ailanthus altissima S., M-B) and a mixed conifer-broadleaf (Robinia-Pinus, M-CB) stand. Soil physicochemical properties, C mineralization potential, enzyme activities, and the abundance of C and nitrogen (N) cycling functional genes were systematically analyzed. Our results showed that the mixed M-CB stand was the most successful model, exhibiting significantly higher soil organic C, total N, and microbial biomass C and N compared to all other models. Intriguingly, this C-rich stand also displayed the highest potential C mineralization rates and enzyme activities (beta-1,4-glucosidase, Alkaline phosphatase), supporting a "high-input, high-turnover" ecosystem dynamic. High-throughput qPCR analysis provided a direct mechanistic link, revealing that the M-CB soil harbored the greatest abundance of genes for both C degradation (e.g., starch, lignin breakdown) and C sequestration. The abundance of these C-degrading genes was strongly and positively correlated with measured C mineralization rates. Overall, this research provides a gene-level mechanistic basis for understanding how plant-plant interactions shape belowground biogeochemical processes and we conclude that the synergistic effects within the Robinia-Pinus stand accelerate the restoration of a functionally active and C-rich soil system.